Isaac D. Gerg, Ph.D.

I develop and rigorously evaluate generative and predictive models for complex, noisy sensor data: diffusion and GAN-based image restoration and synthesis, self-supervised representation learning, learned compression, and the signal processing that makes them work. I do my best work heads-down on a hard technical problem: building an approach, testing it rigorously, understanding why it fails, and charting the next direction. A single thread runs through this work: I build the physics and domain structure of the problem into the model, scattering behavior, coherent-imaging constraints, and signal priors, rather than relying on scale alone. Where I am headed: building frontier generative and world models, physics-based and multimodal learned simulators that fuse real sensor observations with physical structure into representations you can probe, stress-test, and build on.

Before that I spent over a decade pushing deep learning into one of the hardest sensing domains there is: complex-valued, low-data synthetic aperture sonar. Along the way I published 50+ papers, earned two best-paper finalists, and led multi-year research programs and set their technical direction. I earned my Ph.D. at Penn State while working full time.

Open to remote research roles building frontier generative and world models, and advancing the methods behind them: generative modeling, physics-based and domain-enriched ML, and multimodal sensing. Most interested in developing and rigorously evaluating new methods and setting technical direction. US-based, fully remote.